The LED time dependence is now being systematically tracked in the PSU code.

 The time dependence of a problem cell  (NSTB=1, row=10,col=14) is shown below for day 80 data.

The first 8 frames are time dependent LED signals (groups of 40 led events) from day 80.

The lower right frame shows the trend over these 8 runs.

For example, one channel in Small South (row18 col7) is plotted below with a point plotted for every 40 LED events, (~1 min). 

The x axis is day number.

Notice that there is a drop of about 20% from day 80 to day 95.

The question is: Does this represent a variation of LED pulse or a variation of gain?

I address this by looking at the average adc value for a channel on day 80 and day 96 and compare to the LED value for the same two days.

The following plot shows a comparison between two runs  (12080001 12095006)

 The x axis is the ratio of Day 95 to Day 80 LED

 The y axis is the ratio of Day 95 to Day 80 Average ADC

The first plot of shows all channels for NSTB=4 where no short term LED drift is seen

The second plot is for NSTB=1 where short term LED drift is seen

 Conclusion: Little of the large drift in LED pulse height is due to real gain changes!